A Mesa Fix Lands To Take Care Of The R9 290 Issue, Intel/Radeon Performance Problems

[dungeon]

For example this CSGO smoke i would describe as an elephant in the room somewhere and no one seems to know where the cap is

[dungeon]

It is not weird but common, i think you can hear everywhere - bound, idle, cap, bottleneck... or even elephant in the room

[smitty3268]

Generally we always have two parts of the story to look at so GPU and CPU usage. So if one is improved but another stay where it is that is improvment for it, but if another does not stay where it is, reason for improvment is because another was bound.

Yeah, it's just a weird way of describing things at times.

In this case, a CPU stall was removed. Meaning that afterwards the CPU is actually busier, because it's not constantly paused waiting on stuff to finish.

[dungeon]

Generally we always have two parts of the story to look at so GPU and CPU usage. So if one is improved but another stay where it is that is improvment for it, but if another does not stay where it is, reason for improvment is because another was bound.

[smitty3268]

I guess with the patch glxgears is not anymore CPU bound

It's still "CPU bound", whatever that means - it just now can draw on an extra back buffer while the previous one is busy getting displayed. So if anything, it's more CPU bound than before, when the cpu use was getting delayed.

[dungeon]

And what was CPU usage while running glxgears before and after the fix?

[droste]

Edit: 20%

[droste]

50% is well within glxgears margin, 10x is within glxgears margin. Just move the window around and you can see that for yourself. This isn't news.

Ok just did it myself, moved the window to 10 different positions on the monitor.

Before the fix:

15989 frames in 5.0 seconds = 3197.556 FPS
15383 frames in 5.0 seconds = 3076.527 FPS
16739 frames in 5.0 seconds = 3347.562 FPS
18079 frames in 5.0 seconds = 3615.760 FPS
18443 frames in 5.0 seconds = 3688.559 FPS
17479 frames in 5.0 seconds = 3495.794 FPS
16041 frames in 5.0 seconds = 3208.020 FPS
17627 frames in 5.0 seconds = 3525.228 FPS
17963 frames in 5.0 seconds = 3592.569 FPS
18198 frames in 5.0 seconds = 3639.544 FPS
15711 frames in 5.0 seconds = 3142.071 FPS

After the fix:

31419 frames in 5.0 seconds = 6283.713 FPS
32832 frames in 5.0 seconds = 6566.313 FPS
33767 frames in 5.0 seconds = 6753.347 FPS
32371 frames in 5.0 seconds = 6474.170 FPS
27941 frames in 5.0 seconds = 5588.111 FPS
32770 frames in 5.0 seconds = 6553.980 FPS
33574 frames in 5.0 seconds = 6714.664 FPS
29995 frames in 5.0 seconds = 5998.774 FPS
32793 frames in 5.0 seconds = 6558.516 FPS
31121 frames in 5.0 seconds = 6224.073 FPS

That is a solid 100% more FPS and moving the window causes only up to 2% change.

[dungeon]

You have games benchmarked on Phoronix more then anything else .

But there are other things that matters for users, i am sure users would like to see how new glamor behave... as it is 2D accel that matters even more then games for most

[duby229]

EDIT: Games..... That's what we need to see benchmarked....